explaining trade flows in renewable energy …...technological developments in the renewable energy...
TRANSCRIPT
Explaining Trade Flows in Renewable Energy Products: The Role of Technological
Development
Mai MiyamotoKenji Takeuchi
May 2018
Discussion Paper No.1819
GRADUATE SCHOOL OF ECONOMICS
KOBE UNIVERSITY
ROKKO, KOBE, JAPAN
1
Explaining Trade Flows in Renewable Energy Products:
The Role of Technological Development
Mai Miyamoto* and Kenji Takeuchi†
May 10, 2018
Abstract
This study investigates the trade flows of renewable energy products, focusing on the role of
technological development. We estimate a gravity model that explains the trade flows among 35
OECD countries from 1996 to 2010 using patent counts as a proxy for technology level. We
compare the pattern of the trade flows between two representative renewable energy products:
those related to wind and solar electricity generation. The results suggest that technological lev-
el is correlated with trade flows and this correlation is weaker in the model for solar products
than that for wind products. When we include China in the sample in estimation, the technolog-
ical level of solar energy is no longer correlated with the exports of solar power products.
Key Words: Renewable energy products; Trade; Technological development; Patent; Gravity
model
JEL Classification Numbers: F14, O33, Q55
* Graduate School of Economics, Kobe University. 2-1, Rokkodaicho, Kobe, Hyogo, 657-8501 Japan. E-mail: [email protected] † Graduate School of Economics, Kobe University. 2-1, Rokkodaicho, Kobe, Hyogo, 657-8501 Japan. E-mail: [email protected]
2
1. Introduction
As of 2016, renewable power accounted for nearly 62% of the net additions to global power
generating capacity (REN21, 2017). The growth and geographical expansion of renewable en-
ergy capacity was driven by the continued decline in the cost of renewable energy generation.
Technological developments in the renewable energy sector continue to offer the potential for
additional cost reductions. For example, since 2012, the cost of lithium ion battery modules has
declined more than 70% (Clover, 2017). Big data and artificial intelligence can improve weather
forecasting, and blockchain technology can help integrate renewable energy into the grid
(Motyka, 2018). Further technological development is a key driver for the promotion of renew-
able energy by serving as a means to ensure a dramatic decrease in the marginal cost of power
generation, to overcome intermittency, and to stabilize the electricity supply.
There are various channels of economic impacts by technological development in the
renewable energy sector. First, the process of technology improvement reduces the cost of elec-
tricity generation or the cost of the improvement of productivity in the sector (Tang and Popp,
2011). A lighter wind turbine blade or a more efficient solar panel can increase the efficiency of
converting renewable energy to electricity. Second, eventually, innovations in renewable energy
3
technology will lead to the increased use of renewables for electricity generation (Popp et al.,
2011). As renewable energy gains cost advantages, it becomes more attractive to adopt renewa-
bles that can produce electricity with lower carbon emissions. Third, technological development
of the renewable energy sector will increase the exports of products related to electricity genera-
tion (Groba, 2014). Therefore, the accumulation of knowledge in environmental technology
should positively affect the bilateral trade between countries.
This paper focuses on the last channel mentioned above and examines the relation-
ship between technological development and the exports of renewable energy products. The
effects of new innovations are not limited to the country of origin, but spread through the pur-
chase and sale of products embedding the new technology. Therefore, the expansion of exports
can be interpreted as the international spillovers of environmental technology. The proliferation
of technology plays an indispensable role in climate change mitigation, which requires coopera-
tion across the globe.
The aims of this study are threefold. First, we estimate a gravity model to explain
trade flows among 35 OECD countries from 1996 to 2010 by considering patent counts as a
proxy for technology level. Several studies on the determinants of technological development in
4
the renewable energy sector have used patent counts as a proxy for a country’s level of techno-
logical development (Johnstone et al., 2010; Ayari et al., 2012; Popp, et al., 2011; and Rexhäu-
ser and Löschel, 2015). Second, we compare the pattern of trade flows between technologies for
two renewable energy sources: wind and solar. Given the characteristics of the products for
these two renewable energy types, we expect that the effects of technology development on ex-
port values will vary between these two energy sources. Finally, we estimate the effects of re-
newable energy policies on trade flows. Policies promoting renewable energy are broadly di-
vided into international and domestic. This study examines the effect of an international agree-
ment (Kyoto Protocol) and two domestic policy instruments [feed-in-tariff (FIT) and renewable
portfolio standard (RPS)] on trade performance.
The rest of the paper is organized as follows. The next section reviews the literature
on trade flows and technological development. Section 3 explains the models and the data used
in the analysis, and Section 4 presents the results of the empirical study. In Section 5, we esti-
mate the model including China in sample. The last section concludes with a summary of the
main results.
5
2. Literature review
Several economists have investigated the interactions between trade flows and technological
development with regard to environmental technologies. The prevailing motivation for these
studies is the Porter hypothesis that environmental regulations will positively affect innovation
and comparative advantage on the global market (Porter and van der Linde, 1995). Using a
gravity model, Costantini and Crespi (2008) analyzed the determinants of and transmission
channels through which environmental technologies are exported to advanced and developing
countries. Their results were consistent with the Porter hypothesis: stricter environmental regu-
lations, supplemented by a strong national innovation system, were a crucial driver of export
performance in the field of energy technologies. Costantini and Mazzanti (2012) explored how
the export competitiveness of the European Union (EU) has been affected by environmental
regulations and innovation in the context of the Porter hypothesis. They adopted a gravity mod-
el and tested both strong and narrowly strong versions of the Porter hypothesis.1 Although
some differences emerged between the results for those two versions, overall, they found that
both public policies and private innovation increase efficiency in the production process through 1 The strong version of Porter hypothesis claims that environmental regulation will enhance economic performance for compliant firms, the regulated sector, and the economy as a whole. The narrowly strong version posits that a more stringent regulatory framework will positively affect the domestic environmental industry. On the other hand, the weak version predicts that environmental regulations will induce additional innovation.
6
various complementarity mechanisms. Groba (2014) focused on solar energy technology com-
ponents and explored the structure and development of international trade. He estimated the
impact of policy instruments on the trade flows among 21 OECD exporting countries and 118
importing countries between 1999 and 2007. The results indicated that the Porter hypothesis
was valid, with the early adopters of renewable energy policies gaining a comparative ad-
vantage.
This study is different from the other seminal works mentioned above because it ad-
dresses specific renewable energy sectors. Although Costantini and Crespi (2008) and Costan-
tini and Mazzanti (2012) investigated the environmental sector in general, their analyses do not
capture the different roles that technology or policy play in trade of specific renewable energy
sectors in detail. The difference between our study and that of Groba (2014) is that we compare
solar and wind energy technology and use patent counts as a measure of technology develop-
ment.
Our study further contributes to the literature by considering the effects of the im-
porter country’s technology level. Generally, the gravity model explains trade flows in terms of
a bilateral relationship between countries. In this model, the trade flows are based on both the
7
economic size and the distance between exporters and importers. It is highly likely that policies
aiming to promote renewable energy as well as the technological level of exporters and import-
ers will both affect trade flows. This study examines the impacts of technological level of ex-
porters and importers on export values. Thus, we can examine the following two hypotheses: (1)
a higher technological level in terms of renewable energy increases the export value of renewa-
ble energy products from that country; and (2) a lower technological level of renewable energy
leads an increase in the imports of renewable energy products by that country.
8
3. Model and Data
3.1. Gravity model
The gravity model explains trade flows among countries by the market size, which is measured
by GDP and distance between countries (Tinbergen, 1962). Our dataset consists of a panel of 35
OECD countries2 covering the years 1996–2010. Estimates of the log-linear form of the gravity
equation are biased and inconsistent because of heteroscedasticity. Therefore, we use the Pois-
son specification of the trade gravity model, suggested by Santos Silva and Tenreyro (2006). To
mitigate potential endogeneity for variables such as technological level, policy, and import tar-
iffs, their lagged values are used. Furthermore, the exact definition of the lag structure for envi-
ronmental and innovation-independent variables is based on both theoretical assumptions and
empirical findings (Costantini and Mazzanti, 2012). The estimation model used is as follows:
where i and j = 1, …, 35 index the cross-section unit (i denotes the exporting country and j the
2 The 35 countries are Australia, Austria, Belgium, Canada, Chile, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Israel, Italy, Japan, Korea, Luxembourg, Mexico, the Nether-lands, New Zealand, Norway, Poland, Portugal, Slovak Republic, Slovenia, Spain, Sweden, Switzerland, Turkey, the United Kingdom, and the United States.
Tijt = β0 +β1Massijt +β2 lnDij +β3Controlijt +β4Technologyit−1 +β5Technologyjt−1+β6Policyit−1 +β7Policyjt−1 + di + dj + dt +εijt !(1)
9
importing country), and t = 1996, …, 2010 indexes time. The dependent variable (Tijt) is the
export value for each type of renewable energy. The independent variables include Massijt,
which reflects the total economic size of the exporting and importing countries;3 distance be-
tween trading countries (Dij); knowledge stock in exporting and importing countries (Technolo-
gy); and the Policy variables, that are FIT, RPS, and the Kyoto Protocol dummy for the two
countries. The control variables include import tariff as well as cultural and geographical dis-
tances among countries. Residual variation is captured by the error term ε. To control for coun-
try-specific and year-specific effects, we estimate the model including dummy variables for the
exporter country, importer country, and each year.
3.2 Data
Dependent variables: Export value
The dependent variable in this study is the export data among OECD countries. The trade data
used in this study were obtained from UN Comtrade (version 2016)4 for World Integrated
3 Because these fixed effects might also absorb structural differences between countries, thus, reducing statistical robustness using the country dummy, we use different measures to reflect the economic size of trading partners. We apply a proxy of the impact the countries’ economic mass, indicating that the trade value is greater when the overall economic space is larger:
4 http://wits.worldbank.org Massijt = ln(GDPit +GDPjt ).
10
Trade Solution (WITS), based on the relevant Harmonized System (HS) code version HS 1996.5
As mentioned previously, this study focuses on two types of renewable energy products: wind
and solar energy. We define the coverage of HS codes used to collect trade data based on Groba
and Cao (2015). The consumer price index is used to convert the trade data to 2010 US dollars.
Figure 1 shows the export values of wind and solar power products among the 35 OECD coun-
tries. Over the study period, there was a steady increase in the export values for both wind and
solar energy products. Although the export values increased for both products, the increase in
solar power products was more dramatic. Solar products are traded among a large number of
countries, whereas wind power products are traded among relatively fewer countries.
Independent variables
The gravity model explains the export values by the economic size of the importer and exporter
and the distance between the importer and the exporter. We add two independent variables to
the general gravity model: (1) exporter and importer countries’ technological level regarding
renewable energy, represented by their knowledge stock, and (2) international and domestic
5 HS codes differentiate 6,641 groups of products (in version HS 1996), whereas SITD codes classify 4,203 product groups (in version Rec. 3). HS 1996 defined renewable energy products for the first time.
11
policies that promote renewable energy.
Technological level
The number of patent applications is used as a proxy variable for technological development.
Technological level and innovation can be measured by various indicators, such as R&D ex-
penditure, number of researchers, and patent counts. However, R&D expenditure is an input
rather than an output of innovation. In contrast, patent counts can be classified as an output in-
dicator of innovation activity and are more related to realized innovation. Furthermore, patents
serve as a source of rich information on new technologies because they are systematically
screened using a considerable amount of resources by governments over a long period of time
(Nagaoka et al., 2010). Because differences in patent systems and laws between countries raise
an issue of compatibility, we examine applications submitted to the European Patent Office
(EPO) only. The patent data in this study were obtained from the Worldwide Patent Statistical
Database, PATSTAT (version 2016 autumn). We classify the patent data into each renewable
energy source using International Patent Classification (IPC) codes (World Intellectual Property
12
Organization (WIPO), 2012)6.
Figure 2 shows trends in the patent counts for wind energy and solar energy products
among the 35 OECD countries. Similar to the trend of export values, the patent counts for both
energy products show a clear increase with the patent count of solar energy being higher than
that of wind energy. Table 1 reports total export and import values and patent counts for the top
10 OECD countries from 1996 to 2010. Among the OECD countries, Denmark exports the
largest value of wind energy products. In terms of technological level, Denmark is likely one of
the countries with the highest technology level in wind energy. Regarding solar energy, Ger-
many is the largest exporter of these products among the OECD countries. In addition, Germany
has the highest number of patents for solar energy products. Following Groba and Cao (2015),
we use the following equation to calculate the knowledge stock variable:
Accordingly, the knowledge stock equals the respective stock at time t–1 minus the depreciation
rate δ plus the patent applications at time t. Following the previous studies (Guellec et al., 2004,
6 Appendix 1 presents the list of IPC code for renewable energy technologies.
KTechnologyit = (1−δ)KTechnologyit−1 +Patentit !(2)
13
and Groba and Cao, 2015), a depreciation rate of 15% has been adopted.7
Environmental policies: Kyoto Protocol, FIT, and RPS
We include two types of environmental policies that promote renewable energy—Kyoto Pro-
tocol as an international policy and FIT/RPS as domestic policies—into the gravity model. The
Kyoto Protocol variable is a dummy that describes a country’s level of environmental effort.
The variable takes a value of one if both exporter and importer countries signed the Kyoto Pro-
tocol as Annex I countries.8 Moreover, we include variables regarding the implementation of
FIT and RPS to consider the effect of renewable energy promotion policies in the exporter and
importer countries. Regarding the FIT variable, we use the guaranteed purchase price for elec-
tricity generated by solar or wind energy, and the percentage of electricity required to be gener-
ated by renewable energy (or covered by a certificate of renewable energy) for RPS. We con-
struct the panel data for these domestic policy-related variables from the following databases
and reports: (1) IEA/IRENA Global Renewable Energy Policies and Measures Database from
7 We conducted robustness checks with a depreciation rate of 10% and 20%, and confirmed that the main result is not sensitive to the choice of depreciation rate. 8 Because many countries signed the Protocol in 1998, the Kyoto dummy takes a value of zero from 1996 to 1997 and takes a value of one from 1998 by 2010 for those countries. On the other hand, it takes a value of one from 1998 to 2001 for the United States, which withdrew from the Protocol in 2001.
14
OECD/IEA and IRENA;9 (2) Commission of the European Communities (2008); (3) Renewa-
ble Energy Policy Network for the 21st Century (REN 21, 2010 and 2013); (4) IEA reports on
renewable energy (IEA, 2004, 2008, and 2011); and (5) Energy Policies of IEA Countries for
respective countries. For the United States, FIT and RPS were not introduced at a country level;
however, some states introduced policies to promote renewable energy. Therefore, we used
state-level electricity consumption to calculate the national weighted average value of FIT and
RPS.
Other variables
The gravity model assumes that trade flows are proportional to variables such as economy size
(GDP of each exporting country and importing country), distance between trading countries,
and control variables related to trade, such as language. GDP data are obtained from OECD sta-
tistics (version 2015). We use GDP per capita constant PPP in adjusted USD aggregated for the
exporter and importer countries to represent the economic size of trading countries. Because
transportation costs play a significant role in determining trade flows, we include the variable
9 https://www.iea.org/policiesandmeasures/renewableenergy/
15
Distance to represent the geographical distance between the exporter and importer countries
The distance data between exporters and importers are taken from the Centre d’Etudes Prospec-
tives et d’Informations Internationales (CPEII) GeoDist dataset. The distance between the capi-
tal cities of each country is used as a proxy for trade costs. We use the import tariff applied to
the products of solar and wind energy by the importing country to reflect the actual levels of
trade barriers. Import tariff data are obtained from the UNCTAD TRAINS10 database using the
HS codes related to solar and wind energy products. Furthermore, to control for the impact of
cultural and geographical similarity on trade flows, we include three types of dummy variables:
language, contiguity, and common currency. The language variable takes the value of one for a
common primary official language between the exporter and the importer. The contiguity varia-
ble takes the value of one for an exporter/ importer pair that shares a country border. Similarly,
the common currency variable takes the value of one when the exporter and the importer use a
common currency. These data are from the CPEII’s gravity dataset (Head et al., 2010; Mayer
and Zignago, 2011). Table 2 provides summary statistics of these variables. We verified corre-
lation among explanatory variables and found it to be below 0.4.
10 http://wits.worldbank.org
16
4. Results
Table 3 presents the main results. This paper mainly discusses the results of the PPML model in
columns (1) and (2). In addition, we show the results of the Negative Binomial (NB) model and
OLS regression in specifications (3), (4), (5), and (6). All the models include exporter country,
importer country, and year dummies to control for country and year effects.
Regarding wind energy, the coefficients on Technological Level of exporter countries
are positive and statistically significant at the 1% level. These coefficients indicate that a
one-point11 increase in the technological level of wind energy in an exporter countries raises
that country’s trade value by approximately 1.4%. In other words, it indicates that a 1,000 count
increase in patent applications will raise trade value by approximately 8 million USD.12 Re-
garding the solar energy, coefficients on technological level are positive and statistically signif-
icant at the 5% level in the PPML model. This finding means that a one-point increase in the
technological level of solar energy in exporter countries will raise export values by approxi-
mately 0.3%. In other words, a 1,000 increase in the patent count will raise trade value by ap-
proximately 5 million USD. These results suggest that exporting countries with a higher tech-
11 In this case, one point of technological level is 1,000 patents. 12 The calculation is based on the mean export value.
17
nological level will have increased export values. Therefore, hypothesis (1)—a higher level of
technological development in an exporter country can increase the export value of renewable
energy products—is supported.
When we look at the importer side, the coefficient of technological level for wind
energy is negative and statistically significant. These coefficients indicate that one-point in-
crease in the technological level of wind energy in the importer countries will reduce trade value
by approximately 1.4%. In other words, a 100 decrease in patent counts in the importer coun-
tries will increase exports to these countries by approximately 0.8 million USD. These results
suggest that the importing countries having a lower technology level for wind energy technolo-
gy will positively affect their import value. Therefore, hypothesis (2) is supported: a lower
technological level in the importing country might increase the import value of wind energy
products. For solar energy, on the other hand, we do not find a statistically significant effect of
technological level for importer countries in all models.
Comparing the effects of solar and wind technology on export values, we find some
differences in the estimated coefficients. The technological level of an importer is statistically
significant only for the export of wind energy products. In the case of solar energy, importer
18
technological level does not affect trade value. On the other hand, with regard to the technolog-
ical level of exporter countries, statistically significant impacts exist for both energy types. Re-
garding the size of the technology effect between the two studied renewable energy sources, the
coefficient for wind is relatively larger. This result shows that a stronger relationship exists be-
tween the technological level of the exporters and the trade flows in the case of wind energy
than for solar energy.
Estimation results for policy related variables are somewhat mixed, and statistical
significance varies depending on energy source and policy instruments. The coefficient for the
Kyoto Protocol is positive and statistically significant at the 1% level for solar energy; however,
the coefficient is not significant for wind energy. The coefficient of the importer being a signa-
tory of Kyoto is positive and statistically significant at the 1% level for solar. However, for
wind energy, that variable has no statistically significant effect on export values. These results
suggest that the Kyoto Protocol has a positive impact on trade flows for solar energy products.
The size of the coefficients suggest that countries that sign the Kyoto Protocol see import values
rise by approximately 0.5%. In other words, these countries raise their trade value by approxi-
mately 10 million USD. For wind energy, the effect of signing Kyoto is observed on the export-
19
er side only.
When we look at the effects of domestic policy instruments, the coefficient on FIT in
importer is positive and statistically significant at the 1% level for solar. The coefficient on RPS
in exporters is negative, whereas that in importers is positive for wind energy. The coefficients
on RPS in exporters and importers are negative for solar energy. Therefore, regarding the RPS
variable, our estimation results do not support the Porter hypothesis.
The estimated coefficients for the other variables typically included in gravity model
are in line with our expectations. For example, the coefficient of Mass is positive and significant
at the 1% level for solar energy. This result demonstrate that the economy size of both the ex-
porter and the importer has a positive impact on export values. The coefficient on Distance is
negative and significant at the 1% level for both types of energy source, suggesting that the dis-
tance between trading countries has a negative impact on export values. The coefficient of
Language is positive and significant at the 1% level for wind energy and at the 5% level for so-
lar energy. The coefficient of Contiguity is positive and significant at the 1% level for wind en-
ergy and at the 5% level for solar energy. The coefficient of Common currency is positive and
statistically significant the 1% level for solar and at the 10% level for wind. The coefficient of
20
Import tariff is negative and significant at the 5% level for solar, suggesting that, as expected
import tariffs have a negative impact on export values for solar energy products. However, tar-
iffs are not statistically significant for wind energy.
21
5. Extensions
5.1 Exports from China
The estimations presented in the previous section used a dataset of trade among 35 OECD
countries. In this subsection, we estimate the same model using a dataset that includes China.
China has dramatically increased the production of renewable energy products in recent years.
In particular, China is a major exporter of solar energy products such as solar panels. Figures 3
and 4 show trends in export values for wind and solar energy products for all the OECD coun-
tries and for China. These figures suggest that solar energy product exports from China in-
creased remarkably after 2005. The export values of solar energy products from China reached
around half the total export values of the 35 analyzed OECD countries in 2010.
The literature dealing with China’s impact on the trade of renewable energy products
is limited. For example, Groba and Cao (2015) investigated the exports of solar and wind ener-
gy technology components from China. They estimate a panel data model on the annual bilat-
eral trade flows of 43 countries that imported solar PV and wind energy products from China
between 1996 and 2008. Their results suggest that the Chinese PV industry entered foreign
markets successfully, although the domestic market remained underdeveloped. Our analysis in
22
this subsection differs from that of Groba and Cao (2015) in that we focus on the trade among
OECD countries and China and the impact of renewable energy technology on the trade values.
Table 4 shows the estimation results using the dataset including China. The specifi-
cation in column (1) for wind energy products indicates that the coefficient for exporter techno-
logical level is positive and that for importer is negative and both are statistically significant at
the 1% level. These results support our hypotheses on the relationship between technological
level and trade values. In contrast, we do not find a significant effect from technological level
for solar energy products. Taken together with the results of Table 3, the effect of technological
level for solar energy is less related to product trade values.
Despite the fact that China exports substantial amounts of solar energy products, it
has fewer patent counts compared with OECD countries.13 The total export value from China to
OECD countries from 1996 to 2010 was 98 billion USD, the largest among the 36 countries
involved in this study. For solar energy products, Chinese companies only registered 116 pa-
tents from 1996 to 2010, fewer than one-tenth of the number of patents registered in Japan, the
US, or Germany.
13 China’s domestic patent counts have increased in recent years. Dang and Motohashi (2015) claim that a subsidy for the filing fee generated applications of lower quality of patents in China.
23
5.2 PCT application
The analysis in Section 4 used patent applications to EPO as a variable indicating technological
level.14 As explained in Section 3, we use EPO applications because this allows us to avoid the
effect caused by different patent laws and systems among countries. In addition, the EPO data
are superior to data from national patent offices because cost differences serves as a quality
hurdle that can eliminate applications for low-value inventions (Johnstone et al., 2010). How-
ever, the data do not capture the patent that is applied to outside European countries. In this
subsection, we re-estimate the model using PCT applications instead of EPO applications. A
PCT application is used for filing a patent abroad. Some literatures indicate that patent filings
abroad are qualitatively higher-grade than patent filings in a domestic patent office (Lanjouw
and Mody, 1996; Lanjouw and Schankerman, 2004; Popp et al., 2011).
Table 5 presents the results using the number of PCT application as the variable in-
dicating technological level. Regarding wind energy technology, the coefficient for technologi-
cal level of exporter is positive but that of importer is negative. Both coefficients are statistically
significant at the 1% level. On the other hand, the coefficient of exporter technological level is 14 The EPO application protect patent rights in European countries. It allows patent rights to be obtained in any one or more of the European Patent Convention (EPC) contracting states by making a single European patent application to the EPO. This may be considerably cheaper than making a separate national application in each EPC member country.
24
not statistically significant but that for importer is negative and statistically significant, both at
the 1% level. These results again suggest that technological level has a clearly observable effect
on trade values in wind energy technology, but less so in solar energy technology.
25
6. Conclusion
This study has investigated the trade flows in renewable energy products, focusing on the role
of technological development. We estimated a gravity model that included knowledge stocks
measured using patent count data as a proxy for the technological level of exporting and im-
porting countries. We compared the pattern of trade flows between two renewable energy tech-
nologies: wind and solar.
The findings in this study highlight three points regarding the development of re-
newable energy technology. First, the technological level of both exporters and importers affect
the trade flows in wind energy products. These findings support our two hypotheses regarding
the relationship between technological level and export values. Second, differences exist in
trade patterns between wind and solar energy products. Specifically, the relationship between
technological level and export values cannot be confirmed in many models for solar energy
products. This result might be due to the different characteristics of wind and solar power tech-
nologies. For example, the production technology of solar power cells uses module manufac-
turing equipment and assembly lines. Even if a firm does not have the requisite knowledge
stock, it can buy turn-key manufacturing equipment and start producing and exporting solar
26
power products. In contrast, wind energy products does not entail such characterstics. Last,
signing the Kyoto protocol has a significant impact on increasing trade flows. The impact of
domestic policies such as FIT and RPS are less significant because these may not cause interna-
tional diffusions of new technologies.
Acknowledgement
This work was supported by JSPS KAKENHI Grant Number JP16H03006, JP26241033. Earlier
versions of this paper were presented at the Spring Meeting of Japanese Economic Association
in 2016, the Sixth Congress of the East Asian Association of Environmental and Resource
Economics, and the Annual Meeting of the Society for Environmental Economics and Policy
Studies in 2016 and 2017. Helpful comments from Kenta, Nakamura, Keisaku Higashida, Tet-
suya Tsurumi, Isao Kamata, Youngho Chang, and the session participants are gratefully
acknowledged.
27
References
Ayari, N., Blazsek, S., and Mendi, P. (2012). Renewable energy innovations in Europe: A dy-
namic panel data approach. Applied Economics, 44(24), 3135–3147.
Clover, I. (2017) Falling lithium-ion battery prices to drive rapid storage uptake. PV Magazine,
August 3, 2017. <https://pv-magazine-usa.com/2017/08/03/falling-lithium-ion-battery-prices
-to-drive-rapid-storage-uptake/ >
Commission of the European Communities. (2008). The support of electricity from renewable
energy sources. Commission Staff Working Document SEC(2008) 57, 23 January 2008.
Costantini, V., and Crespi, F. (2008). Environmental regulation and the export dynamics of en-
ergy technologies. Ecological Economics, 66, 447–460.
Costantini, V., and Mazzanti, M. (2012). On the green and innovative side of trade competi-
tiveness? The impact of environmental policies and innovation on EU exports. Research Pol-
icy, 41, 132–153.
Dang, J., and Motohashi, K. (2015). Patent statistics: A good indicator for innovation in China?
Patent subsidy program impacts on patent quality. China Economic Review, 35, 137_155.
EPO. (2016). World patent statistical database (PATSTAT). European Patent Office, Munich.
28
Groba, F. (2014). Determinants of trade with solar energy technology components: Evidence on
the porter hypothesis? Applied Economics, 46(5), 503–526.
Groba, F., and Cao, J. (2015). Chinese renewable energy technology exports: The role of policy,
innovation and markets. Environmental and Resource Economics, 60, 243–283.
Guellec, D., and van Pottelsberghe de la Potterie, B. (2004). From R&D to productivity growth:
do the institutional settings and the source of funds of R&D matter? Oxford Bulletin of Eco-
nomics and Statistics, 66(3), 353–378
Head, K., Mayer, T., and Ries, J. (2010). The erosion of colonial trade linkages after independ-
ence. Journal of International Economics, 81(1), 1–14.
IEA. (2004). Renewable Energy Market & Policy Trends in IEA Countries. International Ener-
gy Agency, Paris, France.
IEA. (2008). Deploying Renewables Principles for Effective Policies. International Energy
Agency, Paris, France.
IEA. (2011). Deploying Renewables 2011 Best and Future Policy Practice. International Energy
Agency, Paris, France.
IEA/IRENA. Global Renewable Energy Policies and Measures Database from OECD/IEA and
29
IRENA. < https://www.iea.org/policiesandmeasures/renewableenergy/>
Johnstone, N., Haščič, I., and Popp, D. (2010). Renewable energy policies and technological
innovation: Evidence based on patent counts. Environmental and Resource Economics, 45,
133–155.
Lanjouw, J.O., and Mody, A. (1996). Innovation and the international diffusion of environmen-
tally responsive technology, Research Policy, 25, 549–571.
Lanjouw, J.O., and Schankerman, M. (2004). The quality of ideas: measuring innovation with
multiple indicators, Economic Journal, 114 (495), 441–465.
Mayer, T., and Zignago, S. (2011). Notes on CPEII’s distances measures: The Geo Dist Da-
ta-base. CPEII. Working Paper, 2011–25.
Motyka, M. (2018). 2018 outlook on renewable energy. Deloitte Center for Energy Solutions.
Nagaoka, S., Motohashi, K., and Goto, A. (2010). Patent Statistics as An Innovation Indicator.
Handbook of the Economics of Innovation, Volume 2. 1083–1127.
Popp, D., Haščič, I., and Medhi, N. (2011). Technology and diffusion of renewable energy. En-
ergy Economics, 3, 648–662.
Porter, M.E., and van der Linde, C. (1995) Toward a new conception of the environ-
30
ment-competitiveness relationship. Journal of Economic Perspectives, 9, 497–118.
REN21. (2010). Renewables 2010 Global Status Report. Paris: REN21 Secretariat.
REN21. (2013). Renewables 2013 Global Status Report. Paris: REN21 Secretariat.
REN21. (2017). Renewables 2017 Global Status Report. Paris: REN21 Secretariat.
Rexhäuser, S., and Löschel, A. (2015). Invention in energy technologies: Comparing energy
efficiency and renewable energy inventions at the firm level. Energy Policy, 83, 206–217.
Santos Silva, J.M.C., and Tenreyro, S. (2006). The log of gravity. The Review of Economics
and Statistics, 88(4), 641–658.
Tang, T., and Popp, D. (2016). The learning process and technological change in wind power:
evidence from China's CDM wind projects. Journal of Policy Analysis and Management,
35(1), 195–222.
Tinbergen, J. (1962). Shaping the World Economy Suggestions for an International Economic
Policy. New York: Twentieth Century Fund.
United Nations Statistical Division. (2016). UN Comtrade Database.
<http://wits.worldbank.org>
United Nations Conference on Trade and Development. (2016). UCTAD Trade Analysis In-
31
formation System (TRAINS). <http://wits.worldbank.org>
WIPO. (2012). International patent classification (IPC) green inventory. World Intellectual
Property Organization, Geneva.
32
Figure 1. Trade values for wind and solar energy in OECD countries
33
Figure 2. Patent counts for wind and solar energy in OECD countries
34
Figure 3. Export values of wind energy products in OECD countries and China
35
Figure 4. Export values of solar energy products in OECD countries and China
36
Table 1. Top 10 countries for trade values and patent counts: 1996 to 2010
Wind energy
Exporter Export value
(Billion USD) Importer
Import value
(Billion USD) Country Patent counts
1 Denmark 20.1289
US 20.5918
Germany 1669
2 Germany 15.3124
Germany 13.0779
Denmark 832
3 US 14.9372
UK 8.2467
US 649
4 Italy 11.5053
Canada 7.7310
Japan 399
5 Spain 4.8716
France 6.8452
Spain 224
6 Japan 3.8503
Italy 4.0816
UK 218
7 UK 3.8392
Spain 3.9054
Netherlands 167
8 Switzerland 3.4633
Mexico 3.2147
Italy 145
9 Canada 2.9047
Turkey 2.9611
France 127
10 France 2.8755 Japan 2.6910 Sweden 96
Solar energy
Exporter Export value
(Billion USD) Importer
Export value
(Billion USD) Country Patent counts
1 Germany 56.5429
Germany 52.7061
Japan 4637
2 US 46.0080
US 50.1897
US 3564
3 Japan 32.7133
Italy 21.1962
Germany 3276
4 Mexico 20.5662
France 17.0871
France 677
5 UK 19.4950
Spain 15.8746
UK 655
6 Netherlands 14.2741
UK 15.5897
Rep. of Korea 645
7 France 11.8001
Canada 13.8867
Netherlands 415
8 Italy 8.7521
Netherlands 13.2672
Switzerland 382
9 Belgium 7.6609
Rep. of Korea 9.6412
Italy 338
10 Finland 7.5404 Mexico 9.2082 Belgium 197
37
Table 2. Descriptive statistics
Note: Although several variables were constructed for technological level for exporters and importers, the table
shows only one of them because their summary statistics are same. Policy variables are treated similarly. The varia-
bles of import tariff, FIT, and RPS are expressed in hundreds of units, while the technological level is expressed in
thousands.
Obs. Mean Std. Dev. Min Max
Dependent variables: Export value (million US$)
Wind 17850 5.664 30.811 0 1106.368
Solar 17850 16.295 81.930 0 3365.387
Independent variables
Mass ijt ,ln 17850 11.008 0.328 9.541 11.928
Distance ij ,ln 17850 7.988 1.174 4.088 9.883
Language ij 17850 0.074 0.262 0 1
Contiguity ij 17850 0.064 0.245 0 1
Common currency ijt 17850 0.002 0.041 0 1
Import tariff of wind jt-1 17850 0.013 0.024 0 0.153
Import tariff of solar jt-1 17850 0.015 0.023 0 0.120
Technological levelof wind it-1
17850 0.032 0.095 0 0.992
Technological levelof solar it-1 17850 0.125 0.348 0 2.381
Kyoto Protocol it-1 17850 0.726 0.446 0 1
FIT for wind it-1 17850 0.029 0.039 0 0.214
FIT for solar it-1 17850 0.090 0.159 0 0.700
RPS for wind it-1 17850 0.006 0.022 0 0.169
RPS for solar it-1 17850 0.006 0.022 0 0.169
38
Table 3. Main results
(1) (2) (3) (4) (5) (6)wind solar wind solar wind solar0.614 4.121*** 0.253 2.371*** 2.809** 6.713***
(0.900) (1.594) (0.745) (0.744) (1.255) (1.220)-0.734*** -0.876*** -0.962*** -1.082*** -2.085*** -1.943***
(0.068) (0.066) (0.070) (0.063) (0.147) (0.122)0.348*** 0.252** 0.503*** 0.306** 0.501 0.438(0.122) (0.116) (0.160) (0.146) (0.335) (0.285)
0.389*** 0.362** 0.550*** 0.346** 0.308 -0.193(0.149) (0.160) (0.180) (0.163) (0.518) (0.438)0.573* 1.327*** 0.295 0.834*** 0.865 1.126**(0.297) (0.187) (0.235) (0.248) (0.624) (0.557)-0.801 -6.495** 4.967** -2.431* 7.172** 6.999**(1.649) (2.638) (2.450) (1.372) (3.384) (3.031)
1.363*** 0.282** 1.581*** 0.373*** -0.503 -1.455***(0.309) (0.140) (0.403) (0.138) (0.560) (0.242)
-1.434*** 0.025 -0.541 -0.104 -0.365 -0.436(0.360) (0.143) (0.353) (0.143) (0.728) (0.319)0.129 0.587*** 0.386** 0.631*** 0.145 0.961***
(0.101) (0.094) (0.172) (0.121) (0.217) (0.220)0.069 0.588*** 0.217 0.214** 0.075 0.037
(0.150) (0.093) (0.133) (0.094) (0.232) (0.251)-0.411 0.076 -0.590 0.519*** 1.190 0.085(1.358) (0.223) (0.897) (0.178) (1.549) (0.335)-0.764 0.594** -0.146 0.747*** -0.335 0.510(1.514) (0.291) (0.879) (0.192) (1.571) (0.366)
-6.476*** -5.565*** -4.223*** -5.210*** -5.142** -6.061**(2.078) (1.573) (1.431) (1.009) (2.102) (2.442)
7.250*** -4.222** 2.662 -2.242 -4.444** -6.577***(2.417) (1.923) (1.784) (1.802) (2.217) (2.436)1.116 -38.455** 7.023 -15.933* -13.335 -61.157***
(10.161) (17.992) (8.482) (8.380) (14.311) (13.873)Year effects Yes Yes Yes Yes Yes Yes
Exporter effects Yes Yes Yes Yes Yes YesImporter effects Yes Yes Yes Yes Yes Yes
Observations 17850 17850 17850 17850 17850 17850R-squared 0.682 0.748 0.678 0.702
Adjusted R-squared 0.676 0.700AIC 109520.229 188457.000 50166.973 70520.688 95287.531 92379.260BIC 110268.046 189204.817 50914.790 71268.505 96035.348 93127.077
log_likelihood -54664.115 -94132.500 -24987.486 -35164.344 -47547.766 -46093.630
Constant
Contiguity ij
Commoncurrency ijt
Import tariff jt-1
Technologicallevel it-1
Technologicallevel jt-1
KyotoProtocol it-1
KyotoProtocol jt-1
FIT it-1
FIT jt-i
RPS it-1
RPS jt-1
Language ij
PPML OLSNB
Mass ijt, ln
Distance ij, ln
Note: Standard errors are in parentheses. *, **, and *** indicate statistical significance at the p < 0.1, p < 0.05, and p
< 0.01 levels, respectively.
39
Table 4. Results including China
(13) (14) (15) (16) (17) (18)wind solar wind solar wind solar
2.646*** 5.379*** 1.655** 3.244*** 3.665*** 6.789***(0.845) (1.271) (0.644) (0.615) (1.049) (1.011)
-0.738*** -0.773*** -0.954*** -1.070*** -2.032*** -1.915***(0.057) (0.057) (0.062) (0.057) (0.134) (0.110)
0.351*** 0.374*** 0.521*** 0.329** 0.536 0.431(0.122) (0.145) (0.160) (0.145) (0.332) (0.281)
0.376*** 0.412** 0.574*** 0.361** 0.357 -0.168(0.142) (0.173) (0.182) (0.165) (0.525) (0.439)0.602** 1.550*** 0.313 0.892*** 0.955 1.228**(0.300) (0.240) (0.230) (0.250) (0.619) (0.550)-1.049 -5.319*** 4.874** -2.211* 6.328** 5.523**(1.651) (1.925) (2.351) (1.181) (3.202) (2.388)
1.161*** -0.153 1.482*** 0.280** -0.703 -1.496***(0.292) (0.198) (0.398) (0.136) (0.542) (0.233)
-1.336*** -0.046 -0.403 -0.149 -0.378 -0.483(0.360) (0.257) (0.368) (0.145) (0.704) (0.307)-0.001 0.242* 0.174 0.357*** -0.149 0.905***(0.119) (0.132) (0.165) (0.123) (0.214) (0.205)0.036 0.531*** 0.150 0.146 -0.025 -0.015
(0.138) (0.184) (0.131) (0.096) (0.224) (0.234)-0.985 -0.235 -1.530* 0.321* 0.242 0.078(1.208) (0.267) (0.865) (0.176) (1.479) (0.327)-0.764 0.667* -0.529 0.748*** -0.350 0.450(1.340) (0.366) (0.833) (0.198) (1.511) (0.356)
-7.595*** -10.389*** -5.282*** -6.957*** -6.365*** -6.470***(1.984) (2.207) (1.429) (1.052) (2.054) (2.421)
7.186*** -5.625** 2.077 -2.836 -5.179** -6.954***(2.406) (2.399) (1.727) (1.792) (2.178) (2.370)
-21.563** -53.023*** -8.594 -25.534*** -23.148* -62.202***(9.495) (14.603) (7.366) (6.935) (11.985) (11.534)
Year effects Yes Yes Yes Yes Yes YesExporter effects Yes Yes Yes Yes Yes YesImporter effects Yes Yes Yes Yes Yes Yes
Observations 18900 18900 18900 18900 18900 18900R-squared 0.680 0.692 0.680 0.708
Adjusted R-squared 0.678 0.706AIC 117814.023 275679.144 54425.131 78025.842 100751.433 97520.809BIC 118583.020 276448.141 55194.129 78794.840 101520.431 98289.807
log_likelihood -58809.011 -137741.572 -27114.566 -38914.921 -50277.716 -48662.405
Constant
Technologicallevel it-1
Technologicallevel jt-1
KyotoProtocol it-1
KyotoProtocol jt-1
FIT it-1
FIT jt-i
PPML OLSNB
RPS it-1
RPS jt-1
Import tariff jt-1
Mass ijt, ln
Distance ij, ln
Language ij
Contiguity ij
Commoncurrency ijt
Note: Standard errors are is parentheses. *, **, and *** indicate statistical significance at the p < 0.1, p < 0.05, and p
< 0.01 levels, respectively.
40
Table 5. Results using PCT application data
(7) (8) (9) (10) (11) (12)wind solar wind solar wind solar0.758 3.379** 0.285 2.179*** 2.796** 6.852***
(0.932) (1.540) (0.750) (0.737) (1.270) (1.214)-0.734*** -0.881*** -0.963*** -1.080*** -2.085*** -1.944***
(0.068) (0.066) (0.070) (0.064) (0.147) (0.122)0.347*** 0.239** 0.501*** 0.305** 0.500 0.442(0.122) (0.118) (0.161) (0.146) (0.335) (0.285)
0.383*** 0.377** 0.546*** 0.345** 0.308 -0.192(0.149) (0.160) (0.181) (0.163) (0.518) (0.439)0.576* 1.304*** 0.289 0.826*** 0.864 1.130**(0.297) (0.187) (0.235) (0.248) (0.624) (0.557)-0.801 -6.143** 4.900** -2.322* 7.113** 7.261**(1.685) (2.545) (2.463) (1.385) (3.401) (3.020)
2.476*** -0.058 2.319*** 0.153 -1.068 -0.989***(0.571) (0.108) (0.742) (0.128) (1.022) (0.198)
-2.192*** -0.357*** -0.343 -0.295*** -0.397 -0.446*(0.649) (0.117) (0.651) (0.104) (1.304) (0.252)0.244** 0.506*** 0.445** 0.595*** 0.123 0.945***(0.102) (0.078) (0.177) (0.127) (0.219) (0.223)-0.031 0.372*** 0.210 0.158* 0.069 0.004(0.144) (0.073) (0.133) (0.096) (0.234) (0.256)-1.522 0.096 -1.080 0.539*** 1.222 0.017(1.492) (0.191) (0.902) (0.179) (1.531) (0.337)0.792 0.614** 0.120 0.747*** -0.245 0.490
(1.706) (0.257) (0.870) (0.195) (1.559) (0.367)-7.657*** -5.838*** -4.885*** -5.241*** -4.838** -5.785**
(2.194) (1.725) (1.459) (1.012) (2.159) (2.491)9.504*** -3.585* 2.769 -1.968 -4.334* -6.242**(2.805) (1.919) (1.792) (1.829) (2.231) (2.473)-0.575 -29.471* 6.593 -13.637 -13.155 -62.631***
(10.537) (17.425) (8.551) (8.312) (14.493) (13.818)Year effects Yes Yes Yes Yes Yes Yes
Exporter effects Yes Yes Yes Yes Yes YesImporter effects Yes Yes Yes Yes Yes Yes
Observations 17850 17850 17850 17850 17850 17850R-squared 0.683 0.750 0.678 0.701
Adjusted R-squared 0.676 0.700AIC 109742.871 187990.984 50183.408 70523.525 95287.365 92402.939BIC 110490.688 188738.801 50931.225 71271.342 96035.182 93150.756
log_likelihood -54775.436 -93899.492 -24995.704 -35165.763 -47547.683 -46105.470
Constant
Technologicallevel it-1
Technologicallevel jt-1
KyotoProtocol it-1
KyotoProtocol jt-1
FIT it-1
FIT jt-i
PPML OLSNB
RPS it-1
RPS jt-1
Import tariff jt-1
Mass ijt, ln
Distance ij, ln
Language ij
Contiguity ij
Commoncurrency ijt
Note: Standard errors are in parentheses. *, **, and *** indicate statistical significance at the p < 0.1, p < 0.05, and p
< 0.01 levels, respectively.
41
Appendix
Table A1. List of IPC code for renewable energy technologies
IPC code ExplanationWind energy
F03D Wind energy
H02K 7/18 Structural association of electric generator with mechanicaldriving motor
B63B 35/00; E04H 12/00; (F03D 11/04) Structural aspects of wind turbinesB60K 16/00 Propulsion of vehicles using wind powerB60L 8/00 Electric propulsion of vehicles using wind powerSolar energy
H01L 27/142, 31/00-31/078;H01G 9/20; H02N 6/00*
H01L 27/30, 51/42–51/48 Using organic materials as the active part
H01L 25/00, 25/03, 25/16, 25/18, 31/042
C01B 33/02; C23C 14/14; C30B 29/06 Silicon; single-crystal growthG05F 1/67 Regulating to the maximum power available from solar cellsF21L 4/00; F21S 9/03 Electric lighting devices with, or rechargeable with, solar cellsH02J 7/35 Charging batteriesH01G 9/20; H01M 14/00 Dye-sensitised solar cells (DSSC)B63H 13/00 Propulsion of marine vessels by wind-powered motors
Devices adapted for the conversion of radiation energy intoelectrical energy
Using organic materials as the active part
Sorce: WIPO(2012)-IPC Green Inventory(http://www.wipo.int/classifications/ipc/en/green_inventory/)
Note: H02N 6/00 was deleted when IPC code is updated in 2014.